]>
git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - block/bio.c
2 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
13 * You should have received a copy of the GNU General Public Licens
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
19 #include <linux/swap.h>
20 #include <linux/bio.h>
21 #include <linux/blkdev.h>
22 #include <linux/uio.h>
23 #include <linux/iocontext.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/kernel.h>
27 #include <linux/export.h>
28 #include <linux/mempool.h>
29 #include <linux/workqueue.h>
30 #include <linux/cgroup.h>
32 #include <trace/events/block.h>
35 * Test patch to inline a certain number of bi_io_vec's inside the bio
36 * itself, to shrink a bio data allocation from two mempool calls to one
38 #define BIO_INLINE_VECS 4
41 * if you change this list, also change bvec_alloc or things will
42 * break badly! cannot be bigger than what you can fit into an
45 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
46 static struct biovec_slab bvec_slabs
[BIOVEC_NR_POOLS
] __read_mostly
= {
47 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES
),
52 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
53 * IO code that does not need private memory pools.
55 struct bio_set
*fs_bio_set
;
56 EXPORT_SYMBOL(fs_bio_set
);
59 * Our slab pool management
62 struct kmem_cache
*slab
;
63 unsigned int slab_ref
;
64 unsigned int slab_size
;
67 static DEFINE_MUTEX(bio_slab_lock
);
68 static struct bio_slab
*bio_slabs
;
69 static unsigned int bio_slab_nr
, bio_slab_max
;
71 static struct kmem_cache
*bio_find_or_create_slab(unsigned int extra_size
)
73 unsigned int sz
= sizeof(struct bio
) + extra_size
;
74 struct kmem_cache
*slab
= NULL
;
75 struct bio_slab
*bslab
, *new_bio_slabs
;
76 unsigned int new_bio_slab_max
;
77 unsigned int i
, entry
= -1;
79 mutex_lock(&bio_slab_lock
);
82 while (i
< bio_slab_nr
) {
83 bslab
= &bio_slabs
[i
];
85 if (!bslab
->slab
&& entry
== -1)
87 else if (bslab
->slab_size
== sz
) {
98 if (bio_slab_nr
== bio_slab_max
&& entry
== -1) {
99 new_bio_slab_max
= bio_slab_max
<< 1;
100 new_bio_slabs
= krealloc(bio_slabs
,
101 new_bio_slab_max
* sizeof(struct bio_slab
),
105 bio_slab_max
= new_bio_slab_max
;
106 bio_slabs
= new_bio_slabs
;
109 entry
= bio_slab_nr
++;
111 bslab
= &bio_slabs
[entry
];
113 snprintf(bslab
->name
, sizeof(bslab
->name
), "bio-%d", entry
);
114 slab
= kmem_cache_create(bslab
->name
, sz
, ARCH_KMALLOC_MINALIGN
,
115 SLAB_HWCACHE_ALIGN
, NULL
);
121 bslab
->slab_size
= sz
;
123 mutex_unlock(&bio_slab_lock
);
127 static void bio_put_slab(struct bio_set
*bs
)
129 struct bio_slab
*bslab
= NULL
;
132 mutex_lock(&bio_slab_lock
);
134 for (i
= 0; i
< bio_slab_nr
; i
++) {
135 if (bs
->bio_slab
== bio_slabs
[i
].slab
) {
136 bslab
= &bio_slabs
[i
];
141 if (WARN(!bslab
, KERN_ERR
"bio: unable to find slab!\n"))
144 WARN_ON(!bslab
->slab_ref
);
146 if (--bslab
->slab_ref
)
149 kmem_cache_destroy(bslab
->slab
);
153 mutex_unlock(&bio_slab_lock
);
156 unsigned int bvec_nr_vecs(unsigned short idx
)
158 return bvec_slabs
[idx
].nr_vecs
;
161 void bvec_free(mempool_t
*pool
, struct bio_vec
*bv
, unsigned int idx
)
163 BIO_BUG_ON(idx
>= BIOVEC_NR_POOLS
);
165 if (idx
== BIOVEC_MAX_IDX
)
166 mempool_free(bv
, pool
);
168 struct biovec_slab
*bvs
= bvec_slabs
+ idx
;
170 kmem_cache_free(bvs
->slab
, bv
);
174 struct bio_vec
*bvec_alloc(gfp_t gfp_mask
, int nr
, unsigned long *idx
,
180 * see comment near bvec_array define!
198 case 129 ... BIO_MAX_PAGES
:
206 * idx now points to the pool we want to allocate from. only the
207 * 1-vec entry pool is mempool backed.
209 if (*idx
== BIOVEC_MAX_IDX
) {
211 bvl
= mempool_alloc(pool
, gfp_mask
);
213 struct biovec_slab
*bvs
= bvec_slabs
+ *idx
;
214 gfp_t __gfp_mask
= gfp_mask
& ~(__GFP_WAIT
| __GFP_IO
);
217 * Make this allocation restricted and don't dump info on
218 * allocation failures, since we'll fallback to the mempool
219 * in case of failure.
221 __gfp_mask
|= __GFP_NOMEMALLOC
| __GFP_NORETRY
| __GFP_NOWARN
;
224 * Try a slab allocation. If this fails and __GFP_WAIT
225 * is set, retry with the 1-entry mempool
227 bvl
= kmem_cache_alloc(bvs
->slab
, __gfp_mask
);
228 if (unlikely(!bvl
&& (gfp_mask
& __GFP_WAIT
))) {
229 *idx
= BIOVEC_MAX_IDX
;
237 static void __bio_free(struct bio
*bio
)
239 bio_disassociate_task(bio
);
241 if (bio_integrity(bio
))
242 bio_integrity_free(bio
);
245 static void bio_free(struct bio
*bio
)
247 struct bio_set
*bs
= bio
->bi_pool
;
253 if (bio_flagged(bio
, BIO_OWNS_VEC
))
254 bvec_free(bs
->bvec_pool
, bio
->bi_io_vec
, BIO_POOL_IDX(bio
));
257 * If we have front padding, adjust the bio pointer before freeing
262 mempool_free(p
, bs
->bio_pool
);
264 /* Bio was allocated by bio_kmalloc() */
269 void bio_init(struct bio
*bio
)
271 memset(bio
, 0, sizeof(*bio
));
272 atomic_set(&bio
->__bi_remaining
, 1);
273 atomic_set(&bio
->__bi_cnt
, 1);
275 EXPORT_SYMBOL(bio_init
);
278 * bio_reset - reinitialize a bio
282 * After calling bio_reset(), @bio will be in the same state as a freshly
283 * allocated bio returned bio bio_alloc_bioset() - the only fields that are
284 * preserved are the ones that are initialized by bio_alloc_bioset(). See
285 * comment in struct bio.
287 void bio_reset(struct bio
*bio
)
289 unsigned long flags
= bio
->bi_flags
& (~0UL << BIO_RESET_BITS
);
293 memset(bio
, 0, BIO_RESET_BYTES
);
294 bio
->bi_flags
= flags
;
295 atomic_set(&bio
->__bi_remaining
, 1);
297 EXPORT_SYMBOL(bio_reset
);
299 static void bio_chain_endio(struct bio
*bio
)
301 struct bio
*parent
= bio
->bi_private
;
303 parent
->bi_error
= bio
->bi_error
;
309 * Increment chain count for the bio. Make sure the CHAIN flag update
310 * is visible before the raised count.
312 static inline void bio_inc_remaining(struct bio
*bio
)
314 bio_set_flag(bio
, BIO_CHAIN
);
315 smp_mb__before_atomic();
316 atomic_inc(&bio
->__bi_remaining
);
320 * bio_chain - chain bio completions
321 * @bio: the target bio
322 * @parent: the @bio's parent bio
324 * The caller won't have a bi_end_io called when @bio completes - instead,
325 * @parent's bi_end_io won't be called until both @parent and @bio have
326 * completed; the chained bio will also be freed when it completes.
328 * The caller must not set bi_private or bi_end_io in @bio.
330 void bio_chain(struct bio
*bio
, struct bio
*parent
)
332 BUG_ON(bio
->bi_private
|| bio
->bi_end_io
);
334 bio
->bi_private
= parent
;
335 bio
->bi_end_io
= bio_chain_endio
;
336 bio_inc_remaining(parent
);
338 EXPORT_SYMBOL(bio_chain
);
340 static void bio_alloc_rescue(struct work_struct
*work
)
342 struct bio_set
*bs
= container_of(work
, struct bio_set
, rescue_work
);
346 spin_lock(&bs
->rescue_lock
);
347 bio
= bio_list_pop(&bs
->rescue_list
);
348 spin_unlock(&bs
->rescue_lock
);
353 generic_make_request(bio
);
357 static void punt_bios_to_rescuer(struct bio_set
*bs
)
359 struct bio_list punt
, nopunt
;
363 * In order to guarantee forward progress we must punt only bios that
364 * were allocated from this bio_set; otherwise, if there was a bio on
365 * there for a stacking driver higher up in the stack, processing it
366 * could require allocating bios from this bio_set, and doing that from
367 * our own rescuer would be bad.
369 * Since bio lists are singly linked, pop them all instead of trying to
370 * remove from the middle of the list:
373 bio_list_init(&punt
);
374 bio_list_init(&nopunt
);
376 while ((bio
= bio_list_pop(current
->bio_list
)))
377 bio_list_add(bio
->bi_pool
== bs
? &punt
: &nopunt
, bio
);
379 *current
->bio_list
= nopunt
;
381 spin_lock(&bs
->rescue_lock
);
382 bio_list_merge(&bs
->rescue_list
, &punt
);
383 spin_unlock(&bs
->rescue_lock
);
385 queue_work(bs
->rescue_workqueue
, &bs
->rescue_work
);
389 * bio_alloc_bioset - allocate a bio for I/O
390 * @gfp_mask: the GFP_ mask given to the slab allocator
391 * @nr_iovecs: number of iovecs to pre-allocate
392 * @bs: the bio_set to allocate from.
395 * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is
396 * backed by the @bs's mempool.
398 * When @bs is not NULL, if %__GFP_WAIT is set then bio_alloc will always be
399 * able to allocate a bio. This is due to the mempool guarantees. To make this
400 * work, callers must never allocate more than 1 bio at a time from this pool.
401 * Callers that need to allocate more than 1 bio must always submit the
402 * previously allocated bio for IO before attempting to allocate a new one.
403 * Failure to do so can cause deadlocks under memory pressure.
405 * Note that when running under generic_make_request() (i.e. any block
406 * driver), bios are not submitted until after you return - see the code in
407 * generic_make_request() that converts recursion into iteration, to prevent
410 * This would normally mean allocating multiple bios under
411 * generic_make_request() would be susceptible to deadlocks, but we have
412 * deadlock avoidance code that resubmits any blocked bios from a rescuer
415 * However, we do not guarantee forward progress for allocations from other
416 * mempools. Doing multiple allocations from the same mempool under
417 * generic_make_request() should be avoided - instead, use bio_set's front_pad
418 * for per bio allocations.
421 * Pointer to new bio on success, NULL on failure.
423 struct bio
*bio_alloc_bioset(gfp_t gfp_mask
, int nr_iovecs
, struct bio_set
*bs
)
425 gfp_t saved_gfp
= gfp_mask
;
427 unsigned inline_vecs
;
428 unsigned long idx
= BIO_POOL_NONE
;
429 struct bio_vec
*bvl
= NULL
;
434 if (nr_iovecs
> UIO_MAXIOV
)
437 p
= kmalloc(sizeof(struct bio
) +
438 nr_iovecs
* sizeof(struct bio_vec
),
441 inline_vecs
= nr_iovecs
;
443 /* should not use nobvec bioset for nr_iovecs > 0 */
444 if (WARN_ON_ONCE(!bs
->bvec_pool
&& nr_iovecs
> 0))
447 * generic_make_request() converts recursion to iteration; this
448 * means if we're running beneath it, any bios we allocate and
449 * submit will not be submitted (and thus freed) until after we
452 * This exposes us to a potential deadlock if we allocate
453 * multiple bios from the same bio_set() while running
454 * underneath generic_make_request(). If we were to allocate
455 * multiple bios (say a stacking block driver that was splitting
456 * bios), we would deadlock if we exhausted the mempool's
459 * We solve this, and guarantee forward progress, with a rescuer
460 * workqueue per bio_set. If we go to allocate and there are
461 * bios on current->bio_list, we first try the allocation
462 * without __GFP_WAIT; if that fails, we punt those bios we
463 * would be blocking to the rescuer workqueue before we retry
464 * with the original gfp_flags.
467 if (current
->bio_list
&& !bio_list_empty(current
->bio_list
))
468 gfp_mask
&= ~__GFP_WAIT
;
470 p
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
471 if (!p
&& gfp_mask
!= saved_gfp
) {
472 punt_bios_to_rescuer(bs
);
473 gfp_mask
= saved_gfp
;
474 p
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
477 front_pad
= bs
->front_pad
;
478 inline_vecs
= BIO_INLINE_VECS
;
487 if (nr_iovecs
> inline_vecs
) {
488 bvl
= bvec_alloc(gfp_mask
, nr_iovecs
, &idx
, bs
->bvec_pool
);
489 if (!bvl
&& gfp_mask
!= saved_gfp
) {
490 punt_bios_to_rescuer(bs
);
491 gfp_mask
= saved_gfp
;
492 bvl
= bvec_alloc(gfp_mask
, nr_iovecs
, &idx
, bs
->bvec_pool
);
498 bio_set_flag(bio
, BIO_OWNS_VEC
);
499 } else if (nr_iovecs
) {
500 bvl
= bio
->bi_inline_vecs
;
504 bio
->bi_flags
|= idx
<< BIO_POOL_OFFSET
;
505 bio
->bi_max_vecs
= nr_iovecs
;
506 bio
->bi_io_vec
= bvl
;
510 mempool_free(p
, bs
->bio_pool
);
513 EXPORT_SYMBOL(bio_alloc_bioset
);
515 void zero_fill_bio(struct bio
*bio
)
519 struct bvec_iter iter
;
521 bio_for_each_segment(bv
, bio
, iter
) {
522 char *data
= bvec_kmap_irq(&bv
, &flags
);
523 memset(data
, 0, bv
.bv_len
);
524 flush_dcache_page(bv
.bv_page
);
525 bvec_kunmap_irq(data
, &flags
);
528 EXPORT_SYMBOL(zero_fill_bio
);
531 * bio_put - release a reference to a bio
532 * @bio: bio to release reference to
535 * Put a reference to a &struct bio, either one you have gotten with
536 * bio_alloc, bio_get or bio_clone. The last put of a bio will free it.
538 void bio_put(struct bio
*bio
)
540 if (!bio_flagged(bio
, BIO_REFFED
))
543 BIO_BUG_ON(!atomic_read(&bio
->__bi_cnt
));
548 if (atomic_dec_and_test(&bio
->__bi_cnt
))
552 EXPORT_SYMBOL(bio_put
);
554 inline int bio_phys_segments(struct request_queue
*q
, struct bio
*bio
)
556 if (unlikely(!bio_flagged(bio
, BIO_SEG_VALID
)))
557 blk_recount_segments(q
, bio
);
559 return bio
->bi_phys_segments
;
561 EXPORT_SYMBOL(bio_phys_segments
);
564 * __bio_clone_fast - clone a bio that shares the original bio's biovec
565 * @bio: destination bio
566 * @bio_src: bio to clone
568 * Clone a &bio. Caller will own the returned bio, but not
569 * the actual data it points to. Reference count of returned
572 * Caller must ensure that @bio_src is not freed before @bio.
574 void __bio_clone_fast(struct bio
*bio
, struct bio
*bio_src
)
576 BUG_ON(bio
->bi_pool
&& BIO_POOL_IDX(bio
) != BIO_POOL_NONE
);
579 * most users will be overriding ->bi_bdev with a new target,
580 * so we don't set nor calculate new physical/hw segment counts here
582 bio
->bi_bdev
= bio_src
->bi_bdev
;
583 bio_set_flag(bio
, BIO_CLONED
);
584 bio
->bi_rw
= bio_src
->bi_rw
;
585 bio
->bi_iter
= bio_src
->bi_iter
;
586 bio
->bi_io_vec
= bio_src
->bi_io_vec
;
588 EXPORT_SYMBOL(__bio_clone_fast
);
591 * bio_clone_fast - clone a bio that shares the original bio's biovec
593 * @gfp_mask: allocation priority
594 * @bs: bio_set to allocate from
596 * Like __bio_clone_fast, only also allocates the returned bio
598 struct bio
*bio_clone_fast(struct bio
*bio
, gfp_t gfp_mask
, struct bio_set
*bs
)
602 b
= bio_alloc_bioset(gfp_mask
, 0, bs
);
606 __bio_clone_fast(b
, bio
);
608 if (bio_integrity(bio
)) {
611 ret
= bio_integrity_clone(b
, bio
, gfp_mask
);
621 EXPORT_SYMBOL(bio_clone_fast
);
624 * bio_clone_bioset - clone a bio
625 * @bio_src: bio to clone
626 * @gfp_mask: allocation priority
627 * @bs: bio_set to allocate from
629 * Clone bio. Caller will own the returned bio, but not the actual data it
630 * points to. Reference count of returned bio will be one.
632 struct bio
*bio_clone_bioset(struct bio
*bio_src
, gfp_t gfp_mask
,
635 struct bvec_iter iter
;
640 * Pre immutable biovecs, __bio_clone() used to just do a memcpy from
641 * bio_src->bi_io_vec to bio->bi_io_vec.
643 * We can't do that anymore, because:
645 * - The point of cloning the biovec is to produce a bio with a biovec
646 * the caller can modify: bi_idx and bi_bvec_done should be 0.
648 * - The original bio could've had more than BIO_MAX_PAGES biovecs; if
649 * we tried to clone the whole thing bio_alloc_bioset() would fail.
650 * But the clone should succeed as long as the number of biovecs we
651 * actually need to allocate is fewer than BIO_MAX_PAGES.
653 * - Lastly, bi_vcnt should not be looked at or relied upon by code
654 * that does not own the bio - reason being drivers don't use it for
655 * iterating over the biovec anymore, so expecting it to be kept up
656 * to date (i.e. for clones that share the parent biovec) is just
657 * asking for trouble and would force extra work on
658 * __bio_clone_fast() anyways.
661 bio
= bio_alloc_bioset(gfp_mask
, bio_segments(bio_src
), bs
);
665 bio
->bi_bdev
= bio_src
->bi_bdev
;
666 bio
->bi_rw
= bio_src
->bi_rw
;
667 bio
->bi_iter
.bi_sector
= bio_src
->bi_iter
.bi_sector
;
668 bio
->bi_iter
.bi_size
= bio_src
->bi_iter
.bi_size
;
670 if (bio
->bi_rw
& REQ_DISCARD
)
671 goto integrity_clone
;
673 if (bio
->bi_rw
& REQ_WRITE_SAME
) {
674 bio
->bi_io_vec
[bio
->bi_vcnt
++] = bio_src
->bi_io_vec
[0];
675 goto integrity_clone
;
678 bio_for_each_segment(bv
, bio_src
, iter
)
679 bio
->bi_io_vec
[bio
->bi_vcnt
++] = bv
;
682 if (bio_integrity(bio_src
)) {
685 ret
= bio_integrity_clone(bio
, bio_src
, gfp_mask
);
694 EXPORT_SYMBOL(bio_clone_bioset
);
697 * bio_get_nr_vecs - return approx number of vecs
700 * Return the approximate number of pages we can send to this target.
701 * There's no guarantee that you will be able to fit this number of pages
702 * into a bio, it does not account for dynamic restrictions that vary
705 int bio_get_nr_vecs(struct block_device
*bdev
)
707 struct request_queue
*q
= bdev_get_queue(bdev
);
710 nr_pages
= min_t(unsigned,
711 queue_max_segments(q
),
712 queue_max_sectors(q
) / (PAGE_SIZE
>> 9) + 1);
714 return min_t(unsigned, nr_pages
, BIO_MAX_PAGES
);
717 EXPORT_SYMBOL(bio_get_nr_vecs
);
720 * bio_add_pc_page - attempt to add page to bio
721 * @q: the target queue
722 * @bio: destination bio
724 * @len: vec entry length
725 * @offset: vec entry offset
727 * Attempt to add a page to the bio_vec maplist. This can fail for a
728 * number of reasons, such as the bio being full or target block device
729 * limitations. The target block device must allow bio's up to PAGE_SIZE,
730 * so it is always possible to add a single page to an empty bio.
732 * This should only be used by REQ_PC bios.
734 int bio_add_pc_page(struct request_queue
*q
, struct bio
*bio
, struct page
735 *page
, unsigned int len
, unsigned int offset
)
737 int retried_segments
= 0;
738 struct bio_vec
*bvec
;
741 * cloned bio must not modify vec list
743 if (unlikely(bio_flagged(bio
, BIO_CLONED
)))
746 if (((bio
->bi_iter
.bi_size
+ len
) >> 9) > queue_max_hw_sectors(q
))
750 * For filesystems with a blocksize smaller than the pagesize
751 * we will often be called with the same page as last time and
752 * a consecutive offset. Optimize this special case.
754 if (bio
->bi_vcnt
> 0) {
755 struct bio_vec
*prev
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
757 if (page
== prev
->bv_page
&&
758 offset
== prev
->bv_offset
+ prev
->bv_len
) {
760 bio
->bi_iter
.bi_size
+= len
;
765 * If the queue doesn't support SG gaps and adding this
766 * offset would create a gap, disallow it.
768 if (q
->queue_flags
& (1 << QUEUE_FLAG_SG_GAPS
) &&
769 bvec_gap_to_prev(prev
, offset
))
773 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
777 * setup the new entry, we might clear it again later if we
778 * cannot add the page
780 bvec
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
781 bvec
->bv_page
= page
;
783 bvec
->bv_offset
= offset
;
785 bio
->bi_phys_segments
++;
786 bio
->bi_iter
.bi_size
+= len
;
789 * Perform a recount if the number of segments is greater
790 * than queue_max_segments(q).
793 while (bio
->bi_phys_segments
> queue_max_segments(q
)) {
795 if (retried_segments
)
798 retried_segments
= 1;
799 blk_recount_segments(q
, bio
);
802 /* If we may be able to merge these biovecs, force a recount */
803 if (bio
->bi_vcnt
> 1 && (BIOVEC_PHYS_MERGEABLE(bvec
-1, bvec
)))
804 bio_clear_flag(bio
, BIO_SEG_VALID
);
810 bvec
->bv_page
= NULL
;
814 bio
->bi_iter
.bi_size
-= len
;
815 blk_recount_segments(q
, bio
);
818 EXPORT_SYMBOL(bio_add_pc_page
);
821 * bio_add_page - attempt to add page to bio
822 * @bio: destination bio
824 * @len: vec entry length
825 * @offset: vec entry offset
827 * Attempt to add a page to the bio_vec maplist. This will only fail
828 * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio.
830 int bio_add_page(struct bio
*bio
, struct page
*page
,
831 unsigned int len
, unsigned int offset
)
836 * cloned bio must not modify vec list
838 if (WARN_ON_ONCE(bio_flagged(bio
, BIO_CLONED
)))
842 * For filesystems with a blocksize smaller than the pagesize
843 * we will often be called with the same page as last time and
844 * a consecutive offset. Optimize this special case.
846 if (bio
->bi_vcnt
> 0) {
847 bv
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
849 if (page
== bv
->bv_page
&&
850 offset
== bv
->bv_offset
+ bv
->bv_len
) {
856 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
859 bv
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
862 bv
->bv_offset
= offset
;
866 bio
->bi_iter
.bi_size
+= len
;
869 EXPORT_SYMBOL(bio_add_page
);
871 struct submit_bio_ret
{
872 struct completion event
;
876 static void submit_bio_wait_endio(struct bio
*bio
)
878 struct submit_bio_ret
*ret
= bio
->bi_private
;
880 ret
->error
= bio
->bi_error
;
881 complete(&ret
->event
);
885 * submit_bio_wait - submit a bio, and wait until it completes
886 * @rw: whether to %READ or %WRITE, or maybe to %READA (read ahead)
887 * @bio: The &struct bio which describes the I/O
889 * Simple wrapper around submit_bio(). Returns 0 on success, or the error from
890 * bio_endio() on failure.
892 int submit_bio_wait(int rw
, struct bio
*bio
)
894 struct submit_bio_ret ret
;
897 init_completion(&ret
.event
);
898 bio
->bi_private
= &ret
;
899 bio
->bi_end_io
= submit_bio_wait_endio
;
901 wait_for_completion(&ret
.event
);
905 EXPORT_SYMBOL(submit_bio_wait
);
908 * bio_advance - increment/complete a bio by some number of bytes
909 * @bio: bio to advance
910 * @bytes: number of bytes to complete
912 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
913 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
914 * be updated on the last bvec as well.
916 * @bio will then represent the remaining, uncompleted portion of the io.
918 void bio_advance(struct bio
*bio
, unsigned bytes
)
920 if (bio_integrity(bio
))
921 bio_integrity_advance(bio
, bytes
);
923 bio_advance_iter(bio
, &bio
->bi_iter
, bytes
);
925 EXPORT_SYMBOL(bio_advance
);
928 * bio_alloc_pages - allocates a single page for each bvec in a bio
929 * @bio: bio to allocate pages for
930 * @gfp_mask: flags for allocation
932 * Allocates pages up to @bio->bi_vcnt.
934 * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are
937 int bio_alloc_pages(struct bio
*bio
, gfp_t gfp_mask
)
942 bio_for_each_segment_all(bv
, bio
, i
) {
943 bv
->bv_page
= alloc_page(gfp_mask
);
945 while (--bv
>= bio
->bi_io_vec
)
946 __free_page(bv
->bv_page
);
953 EXPORT_SYMBOL(bio_alloc_pages
);
956 * bio_copy_data - copy contents of data buffers from one chain of bios to
958 * @src: source bio list
959 * @dst: destination bio list
961 * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats
962 * @src and @dst as linked lists of bios.
964 * Stops when it reaches the end of either @src or @dst - that is, copies
965 * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios).
967 void bio_copy_data(struct bio
*dst
, struct bio
*src
)
969 struct bvec_iter src_iter
, dst_iter
;
970 struct bio_vec src_bv
, dst_bv
;
974 src_iter
= src
->bi_iter
;
975 dst_iter
= dst
->bi_iter
;
978 if (!src_iter
.bi_size
) {
983 src_iter
= src
->bi_iter
;
986 if (!dst_iter
.bi_size
) {
991 dst_iter
= dst
->bi_iter
;
994 src_bv
= bio_iter_iovec(src
, src_iter
);
995 dst_bv
= bio_iter_iovec(dst
, dst_iter
);
997 bytes
= min(src_bv
.bv_len
, dst_bv
.bv_len
);
999 src_p
= kmap_atomic(src_bv
.bv_page
);
1000 dst_p
= kmap_atomic(dst_bv
.bv_page
);
1002 memcpy(dst_p
+ dst_bv
.bv_offset
,
1003 src_p
+ src_bv
.bv_offset
,
1006 kunmap_atomic(dst_p
);
1007 kunmap_atomic(src_p
);
1009 bio_advance_iter(src
, &src_iter
, bytes
);
1010 bio_advance_iter(dst
, &dst_iter
, bytes
);
1013 EXPORT_SYMBOL(bio_copy_data
);
1015 struct bio_map_data
{
1017 struct iov_iter iter
;
1021 static struct bio_map_data
*bio_alloc_map_data(unsigned int iov_count
,
1024 if (iov_count
> UIO_MAXIOV
)
1027 return kmalloc(sizeof(struct bio_map_data
) +
1028 sizeof(struct iovec
) * iov_count
, gfp_mask
);
1032 * bio_copy_from_iter - copy all pages from iov_iter to bio
1033 * @bio: The &struct bio which describes the I/O as destination
1034 * @iter: iov_iter as source
1036 * Copy all pages from iov_iter to bio.
1037 * Returns 0 on success, or error on failure.
1039 static int bio_copy_from_iter(struct bio
*bio
, struct iov_iter iter
)
1042 struct bio_vec
*bvec
;
1044 bio_for_each_segment_all(bvec
, bio
, i
) {
1047 ret
= copy_page_from_iter(bvec
->bv_page
,
1052 if (!iov_iter_count(&iter
))
1055 if (ret
< bvec
->bv_len
)
1063 * bio_copy_to_iter - copy all pages from bio to iov_iter
1064 * @bio: The &struct bio which describes the I/O as source
1065 * @iter: iov_iter as destination
1067 * Copy all pages from bio to iov_iter.
1068 * Returns 0 on success, or error on failure.
1070 static int bio_copy_to_iter(struct bio
*bio
, struct iov_iter iter
)
1073 struct bio_vec
*bvec
;
1075 bio_for_each_segment_all(bvec
, bio
, i
) {
1078 ret
= copy_page_to_iter(bvec
->bv_page
,
1083 if (!iov_iter_count(&iter
))
1086 if (ret
< bvec
->bv_len
)
1093 static void bio_free_pages(struct bio
*bio
)
1095 struct bio_vec
*bvec
;
1098 bio_for_each_segment_all(bvec
, bio
, i
)
1099 __free_page(bvec
->bv_page
);
1103 * bio_uncopy_user - finish previously mapped bio
1104 * @bio: bio being terminated
1106 * Free pages allocated from bio_copy_user_iov() and write back data
1107 * to user space in case of a read.
1109 int bio_uncopy_user(struct bio
*bio
)
1111 struct bio_map_data
*bmd
= bio
->bi_private
;
1114 if (!bio_flagged(bio
, BIO_NULL_MAPPED
)) {
1116 * if we're in a workqueue, the request is orphaned, so
1117 * don't copy into a random user address space, just free.
1119 if (current
->mm
&& bio_data_dir(bio
) == READ
)
1120 ret
= bio_copy_to_iter(bio
, bmd
->iter
);
1121 if (bmd
->is_our_pages
)
1122 bio_free_pages(bio
);
1128 EXPORT_SYMBOL(bio_uncopy_user
);
1131 * bio_copy_user_iov - copy user data to bio
1132 * @q: destination block queue
1133 * @map_data: pointer to the rq_map_data holding pages (if necessary)
1134 * @iter: iovec iterator
1135 * @gfp_mask: memory allocation flags
1137 * Prepares and returns a bio for indirect user io, bouncing data
1138 * to/from kernel pages as necessary. Must be paired with
1139 * call bio_uncopy_user() on io completion.
1141 struct bio
*bio_copy_user_iov(struct request_queue
*q
,
1142 struct rq_map_data
*map_data
,
1143 const struct iov_iter
*iter
,
1146 struct bio_map_data
*bmd
;
1151 unsigned int len
= iter
->count
;
1152 unsigned int offset
= map_data
? map_data
->offset
& ~PAGE_MASK
: 0;
1154 for (i
= 0; i
< iter
->nr_segs
; i
++) {
1155 unsigned long uaddr
;
1157 unsigned long start
;
1159 uaddr
= (unsigned long) iter
->iov
[i
].iov_base
;
1160 end
= (uaddr
+ iter
->iov
[i
].iov_len
+ PAGE_SIZE
- 1)
1162 start
= uaddr
>> PAGE_SHIFT
;
1168 return ERR_PTR(-EINVAL
);
1170 nr_pages
+= end
- start
;
1176 bmd
= bio_alloc_map_data(iter
->nr_segs
, gfp_mask
);
1178 return ERR_PTR(-ENOMEM
);
1181 * We need to do a deep copy of the iov_iter including the iovecs.
1182 * The caller provided iov might point to an on-stack or otherwise
1185 bmd
->is_our_pages
= map_data
? 0 : 1;
1186 memcpy(bmd
->iov
, iter
->iov
, sizeof(struct iovec
) * iter
->nr_segs
);
1187 iov_iter_init(&bmd
->iter
, iter
->type
, bmd
->iov
,
1188 iter
->nr_segs
, iter
->count
);
1191 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1195 if (iter
->type
& WRITE
)
1196 bio
->bi_rw
|= REQ_WRITE
;
1201 nr_pages
= 1 << map_data
->page_order
;
1202 i
= map_data
->offset
/ PAGE_SIZE
;
1205 unsigned int bytes
= PAGE_SIZE
;
1213 if (i
== map_data
->nr_entries
* nr_pages
) {
1218 page
= map_data
->pages
[i
/ nr_pages
];
1219 page
+= (i
% nr_pages
);
1223 page
= alloc_page(q
->bounce_gfp
| gfp_mask
);
1230 if (bio_add_pc_page(q
, bio
, page
, bytes
, offset
) < bytes
)
1243 if (((iter
->type
& WRITE
) && (!map_data
|| !map_data
->null_mapped
)) ||
1244 (map_data
&& map_data
->from_user
)) {
1245 ret
= bio_copy_from_iter(bio
, *iter
);
1250 bio
->bi_private
= bmd
;
1254 bio_free_pages(bio
);
1258 return ERR_PTR(ret
);
1262 * bio_map_user_iov - map user iovec into bio
1263 * @q: the struct request_queue for the bio
1264 * @iter: iovec iterator
1265 * @gfp_mask: memory allocation flags
1267 * Map the user space address into a bio suitable for io to a block
1268 * device. Returns an error pointer in case of error.
1270 struct bio
*bio_map_user_iov(struct request_queue
*q
,
1271 const struct iov_iter
*iter
,
1276 struct page
**pages
;
1283 iov_for_each(iov
, i
, *iter
) {
1284 unsigned long uaddr
= (unsigned long) iov
.iov_base
;
1285 unsigned long len
= iov
.iov_len
;
1286 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1287 unsigned long start
= uaddr
>> PAGE_SHIFT
;
1293 return ERR_PTR(-EINVAL
);
1295 nr_pages
+= end
- start
;
1297 * buffer must be aligned to at least hardsector size for now
1299 if (uaddr
& queue_dma_alignment(q
))
1300 return ERR_PTR(-EINVAL
);
1304 return ERR_PTR(-EINVAL
);
1306 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1308 return ERR_PTR(-ENOMEM
);
1311 pages
= kcalloc(nr_pages
, sizeof(struct page
*), gfp_mask
);
1315 iov_for_each(iov
, i
, *iter
) {
1316 unsigned long uaddr
= (unsigned long) iov
.iov_base
;
1317 unsigned long len
= iov
.iov_len
;
1318 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1319 unsigned long start
= uaddr
>> PAGE_SHIFT
;
1320 const int local_nr_pages
= end
- start
;
1321 const int page_limit
= cur_page
+ local_nr_pages
;
1323 ret
= get_user_pages_fast(uaddr
, local_nr_pages
,
1324 (iter
->type
& WRITE
) != WRITE
,
1326 if (ret
< local_nr_pages
) {
1331 offset
= uaddr
& ~PAGE_MASK
;
1332 for (j
= cur_page
; j
< page_limit
; j
++) {
1333 unsigned int bytes
= PAGE_SIZE
- offset
;
1344 if (bio_add_pc_page(q
, bio
, pages
[j
], bytes
, offset
) <
1354 * release the pages we didn't map into the bio, if any
1356 while (j
< page_limit
)
1357 page_cache_release(pages
[j
++]);
1363 * set data direction, and check if mapped pages need bouncing
1365 if (iter
->type
& WRITE
)
1366 bio
->bi_rw
|= REQ_WRITE
;
1368 bio_set_flag(bio
, BIO_USER_MAPPED
);
1371 * subtle -- if __bio_map_user() ended up bouncing a bio,
1372 * it would normally disappear when its bi_end_io is run.
1373 * however, we need it for the unmap, so grab an extra
1380 for (j
= 0; j
< nr_pages
; j
++) {
1383 page_cache_release(pages
[j
]);
1388 return ERR_PTR(ret
);
1391 static void __bio_unmap_user(struct bio
*bio
)
1393 struct bio_vec
*bvec
;
1397 * make sure we dirty pages we wrote to
1399 bio_for_each_segment_all(bvec
, bio
, i
) {
1400 if (bio_data_dir(bio
) == READ
)
1401 set_page_dirty_lock(bvec
->bv_page
);
1403 page_cache_release(bvec
->bv_page
);
1410 * bio_unmap_user - unmap a bio
1411 * @bio: the bio being unmapped
1413 * Unmap a bio previously mapped by bio_map_user(). Must be called with
1414 * a process context.
1416 * bio_unmap_user() may sleep.
1418 void bio_unmap_user(struct bio
*bio
)
1420 __bio_unmap_user(bio
);
1423 EXPORT_SYMBOL(bio_unmap_user
);
1425 static void bio_map_kern_endio(struct bio
*bio
)
1431 * bio_map_kern - map kernel address into bio
1432 * @q: the struct request_queue for the bio
1433 * @data: pointer to buffer to map
1434 * @len: length in bytes
1435 * @gfp_mask: allocation flags for bio allocation
1437 * Map the kernel address into a bio suitable for io to a block
1438 * device. Returns an error pointer in case of error.
1440 struct bio
*bio_map_kern(struct request_queue
*q
, void *data
, unsigned int len
,
1443 unsigned long kaddr
= (unsigned long)data
;
1444 unsigned long end
= (kaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1445 unsigned long start
= kaddr
>> PAGE_SHIFT
;
1446 const int nr_pages
= end
- start
;
1450 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1452 return ERR_PTR(-ENOMEM
);
1454 offset
= offset_in_page(kaddr
);
1455 for (i
= 0; i
< nr_pages
; i
++) {
1456 unsigned int bytes
= PAGE_SIZE
- offset
;
1464 if (bio_add_pc_page(q
, bio
, virt_to_page(data
), bytes
,
1466 /* we don't support partial mappings */
1468 return ERR_PTR(-EINVAL
);
1476 bio
->bi_end_io
= bio_map_kern_endio
;
1479 EXPORT_SYMBOL(bio_map_kern
);
1481 static void bio_copy_kern_endio(struct bio
*bio
)
1483 bio_free_pages(bio
);
1487 static void bio_copy_kern_endio_read(struct bio
*bio
)
1489 char *p
= bio
->bi_private
;
1490 struct bio_vec
*bvec
;
1493 bio_for_each_segment_all(bvec
, bio
, i
) {
1494 memcpy(p
, page_address(bvec
->bv_page
), bvec
->bv_len
);
1498 bio_copy_kern_endio(bio
);
1502 * bio_copy_kern - copy kernel address into bio
1503 * @q: the struct request_queue for the bio
1504 * @data: pointer to buffer to copy
1505 * @len: length in bytes
1506 * @gfp_mask: allocation flags for bio and page allocation
1507 * @reading: data direction is READ
1509 * copy the kernel address into a bio suitable for io to a block
1510 * device. Returns an error pointer in case of error.
1512 struct bio
*bio_copy_kern(struct request_queue
*q
, void *data
, unsigned int len
,
1513 gfp_t gfp_mask
, int reading
)
1515 unsigned long kaddr
= (unsigned long)data
;
1516 unsigned long end
= (kaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1517 unsigned long start
= kaddr
>> PAGE_SHIFT
;
1526 return ERR_PTR(-EINVAL
);
1528 nr_pages
= end
- start
;
1529 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1531 return ERR_PTR(-ENOMEM
);
1535 unsigned int bytes
= PAGE_SIZE
;
1540 page
= alloc_page(q
->bounce_gfp
| gfp_mask
);
1545 memcpy(page_address(page
), p
, bytes
);
1547 if (bio_add_pc_page(q
, bio
, page
, bytes
, 0) < bytes
)
1555 bio
->bi_end_io
= bio_copy_kern_endio_read
;
1556 bio
->bi_private
= data
;
1558 bio
->bi_end_io
= bio_copy_kern_endio
;
1559 bio
->bi_rw
|= REQ_WRITE
;
1565 bio_free_pages(bio
);
1567 return ERR_PTR(-ENOMEM
);
1569 EXPORT_SYMBOL(bio_copy_kern
);
1572 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1573 * for performing direct-IO in BIOs.
1575 * The problem is that we cannot run set_page_dirty() from interrupt context
1576 * because the required locks are not interrupt-safe. So what we can do is to
1577 * mark the pages dirty _before_ performing IO. And in interrupt context,
1578 * check that the pages are still dirty. If so, fine. If not, redirty them
1579 * in process context.
1581 * We special-case compound pages here: normally this means reads into hugetlb
1582 * pages. The logic in here doesn't really work right for compound pages
1583 * because the VM does not uniformly chase down the head page in all cases.
1584 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1585 * handle them at all. So we skip compound pages here at an early stage.
1587 * Note that this code is very hard to test under normal circumstances because
1588 * direct-io pins the pages with get_user_pages(). This makes
1589 * is_page_cache_freeable return false, and the VM will not clean the pages.
1590 * But other code (eg, flusher threads) could clean the pages if they are mapped
1593 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1594 * deferred bio dirtying paths.
1598 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1600 void bio_set_pages_dirty(struct bio
*bio
)
1602 struct bio_vec
*bvec
;
1605 bio_for_each_segment_all(bvec
, bio
, i
) {
1606 struct page
*page
= bvec
->bv_page
;
1608 if (page
&& !PageCompound(page
))
1609 set_page_dirty_lock(page
);
1613 static void bio_release_pages(struct bio
*bio
)
1615 struct bio_vec
*bvec
;
1618 bio_for_each_segment_all(bvec
, bio
, i
) {
1619 struct page
*page
= bvec
->bv_page
;
1627 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1628 * If they are, then fine. If, however, some pages are clean then they must
1629 * have been written out during the direct-IO read. So we take another ref on
1630 * the BIO and the offending pages and re-dirty the pages in process context.
1632 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1633 * here on. It will run one page_cache_release() against each page and will
1634 * run one bio_put() against the BIO.
1637 static void bio_dirty_fn(struct work_struct
*work
);
1639 static DECLARE_WORK(bio_dirty_work
, bio_dirty_fn
);
1640 static DEFINE_SPINLOCK(bio_dirty_lock
);
1641 static struct bio
*bio_dirty_list
;
1644 * This runs in process context
1646 static void bio_dirty_fn(struct work_struct
*work
)
1648 unsigned long flags
;
1651 spin_lock_irqsave(&bio_dirty_lock
, flags
);
1652 bio
= bio_dirty_list
;
1653 bio_dirty_list
= NULL
;
1654 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
1657 struct bio
*next
= bio
->bi_private
;
1659 bio_set_pages_dirty(bio
);
1660 bio_release_pages(bio
);
1666 void bio_check_pages_dirty(struct bio
*bio
)
1668 struct bio_vec
*bvec
;
1669 int nr_clean_pages
= 0;
1672 bio_for_each_segment_all(bvec
, bio
, i
) {
1673 struct page
*page
= bvec
->bv_page
;
1675 if (PageDirty(page
) || PageCompound(page
)) {
1676 page_cache_release(page
);
1677 bvec
->bv_page
= NULL
;
1683 if (nr_clean_pages
) {
1684 unsigned long flags
;
1686 spin_lock_irqsave(&bio_dirty_lock
, flags
);
1687 bio
->bi_private
= bio_dirty_list
;
1688 bio_dirty_list
= bio
;
1689 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
1690 schedule_work(&bio_dirty_work
);
1696 void generic_start_io_acct(int rw
, unsigned long sectors
,
1697 struct hd_struct
*part
)
1699 int cpu
= part_stat_lock();
1701 part_round_stats(cpu
, part
);
1702 part_stat_inc(cpu
, part
, ios
[rw
]);
1703 part_stat_add(cpu
, part
, sectors
[rw
], sectors
);
1704 part_inc_in_flight(part
, rw
);
1708 EXPORT_SYMBOL(generic_start_io_acct
);
1710 void generic_end_io_acct(int rw
, struct hd_struct
*part
,
1711 unsigned long start_time
)
1713 unsigned long duration
= jiffies
- start_time
;
1714 int cpu
= part_stat_lock();
1716 part_stat_add(cpu
, part
, ticks
[rw
], duration
);
1717 part_round_stats(cpu
, part
);
1718 part_dec_in_flight(part
, rw
);
1722 EXPORT_SYMBOL(generic_end_io_acct
);
1724 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1725 void bio_flush_dcache_pages(struct bio
*bi
)
1727 struct bio_vec bvec
;
1728 struct bvec_iter iter
;
1730 bio_for_each_segment(bvec
, bi
, iter
)
1731 flush_dcache_page(bvec
.bv_page
);
1733 EXPORT_SYMBOL(bio_flush_dcache_pages
);
1736 static inline bool bio_remaining_done(struct bio
*bio
)
1739 * If we're not chaining, then ->__bi_remaining is always 1 and
1740 * we always end io on the first invocation.
1742 if (!bio_flagged(bio
, BIO_CHAIN
))
1745 BUG_ON(atomic_read(&bio
->__bi_remaining
) <= 0);
1747 if (atomic_dec_and_test(&bio
->__bi_remaining
)) {
1748 bio_clear_flag(bio
, BIO_CHAIN
);
1756 * bio_endio - end I/O on a bio
1760 * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred
1761 * way to end I/O on a bio. No one should call bi_end_io() directly on a
1762 * bio unless they own it and thus know that it has an end_io function.
1764 void bio_endio(struct bio
*bio
)
1767 if (unlikely(!bio_remaining_done(bio
)))
1771 * Need to have a real endio function for chained bios,
1772 * otherwise various corner cases will break (like stacking
1773 * block devices that save/restore bi_end_io) - however, we want
1774 * to avoid unbounded recursion and blowing the stack. Tail call
1775 * optimization would handle this, but compiling with frame
1776 * pointers also disables gcc's sibling call optimization.
1778 if (bio
->bi_end_io
== bio_chain_endio
) {
1779 struct bio
*parent
= bio
->bi_private
;
1780 parent
->bi_error
= bio
->bi_error
;
1785 bio
->bi_end_io(bio
);
1790 EXPORT_SYMBOL(bio_endio
);
1793 * bio_split - split a bio
1794 * @bio: bio to split
1795 * @sectors: number of sectors to split from the front of @bio
1797 * @bs: bio set to allocate from
1799 * Allocates and returns a new bio which represents @sectors from the start of
1800 * @bio, and updates @bio to represent the remaining sectors.
1802 * The newly allocated bio will point to @bio's bi_io_vec; it is the caller's
1803 * responsibility to ensure that @bio is not freed before the split.
1805 struct bio
*bio_split(struct bio
*bio
, int sectors
,
1806 gfp_t gfp
, struct bio_set
*bs
)
1808 struct bio
*split
= NULL
;
1810 BUG_ON(sectors
<= 0);
1811 BUG_ON(sectors
>= bio_sectors(bio
));
1813 split
= bio_clone_fast(bio
, gfp
, bs
);
1817 split
->bi_iter
.bi_size
= sectors
<< 9;
1819 if (bio_integrity(split
))
1820 bio_integrity_trim(split
, 0, sectors
);
1822 bio_advance(bio
, split
->bi_iter
.bi_size
);
1826 EXPORT_SYMBOL(bio_split
);
1829 * bio_trim - trim a bio
1831 * @offset: number of sectors to trim from the front of @bio
1832 * @size: size we want to trim @bio to, in sectors
1834 void bio_trim(struct bio
*bio
, int offset
, int size
)
1836 /* 'bio' is a cloned bio which we need to trim to match
1837 * the given offset and size.
1841 if (offset
== 0 && size
== bio
->bi_iter
.bi_size
)
1844 bio_clear_flag(bio
, BIO_SEG_VALID
);
1846 bio_advance(bio
, offset
<< 9);
1848 bio
->bi_iter
.bi_size
= size
;
1850 EXPORT_SYMBOL_GPL(bio_trim
);
1853 * create memory pools for biovec's in a bio_set.
1854 * use the global biovec slabs created for general use.
1856 mempool_t
*biovec_create_pool(int pool_entries
)
1858 struct biovec_slab
*bp
= bvec_slabs
+ BIOVEC_MAX_IDX
;
1860 return mempool_create_slab_pool(pool_entries
, bp
->slab
);
1863 void bioset_free(struct bio_set
*bs
)
1865 if (bs
->rescue_workqueue
)
1866 destroy_workqueue(bs
->rescue_workqueue
);
1869 mempool_destroy(bs
->bio_pool
);
1872 mempool_destroy(bs
->bvec_pool
);
1874 bioset_integrity_free(bs
);
1879 EXPORT_SYMBOL(bioset_free
);
1881 static struct bio_set
*__bioset_create(unsigned int pool_size
,
1882 unsigned int front_pad
,
1883 bool create_bvec_pool
)
1885 unsigned int back_pad
= BIO_INLINE_VECS
* sizeof(struct bio_vec
);
1888 bs
= kzalloc(sizeof(*bs
), GFP_KERNEL
);
1892 bs
->front_pad
= front_pad
;
1894 spin_lock_init(&bs
->rescue_lock
);
1895 bio_list_init(&bs
->rescue_list
);
1896 INIT_WORK(&bs
->rescue_work
, bio_alloc_rescue
);
1898 bs
->bio_slab
= bio_find_or_create_slab(front_pad
+ back_pad
);
1899 if (!bs
->bio_slab
) {
1904 bs
->bio_pool
= mempool_create_slab_pool(pool_size
, bs
->bio_slab
);
1908 if (create_bvec_pool
) {
1909 bs
->bvec_pool
= biovec_create_pool(pool_size
);
1914 bs
->rescue_workqueue
= alloc_workqueue("bioset", WQ_MEM_RECLAIM
, 0);
1915 if (!bs
->rescue_workqueue
)
1925 * bioset_create - Create a bio_set
1926 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1927 * @front_pad: Number of bytes to allocate in front of the returned bio
1930 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
1931 * to ask for a number of bytes to be allocated in front of the bio.
1932 * Front pad allocation is useful for embedding the bio inside
1933 * another structure, to avoid allocating extra data to go with the bio.
1934 * Note that the bio must be embedded at the END of that structure always,
1935 * or things will break badly.
1937 struct bio_set
*bioset_create(unsigned int pool_size
, unsigned int front_pad
)
1939 return __bioset_create(pool_size
, front_pad
, true);
1941 EXPORT_SYMBOL(bioset_create
);
1944 * bioset_create_nobvec - Create a bio_set without bio_vec mempool
1945 * @pool_size: Number of bio to cache in the mempool
1946 * @front_pad: Number of bytes to allocate in front of the returned bio
1949 * Same functionality as bioset_create() except that mempool is not
1950 * created for bio_vecs. Saving some memory for bio_clone_fast() users.
1952 struct bio_set
*bioset_create_nobvec(unsigned int pool_size
, unsigned int front_pad
)
1954 return __bioset_create(pool_size
, front_pad
, false);
1956 EXPORT_SYMBOL(bioset_create_nobvec
);
1958 #ifdef CONFIG_BLK_CGROUP
1961 * bio_associate_blkcg - associate a bio with the specified blkcg
1963 * @blkcg_css: css of the blkcg to associate
1965 * Associate @bio with the blkcg specified by @blkcg_css. Block layer will
1966 * treat @bio as if it were issued by a task which belongs to the blkcg.
1968 * This function takes an extra reference of @blkcg_css which will be put
1969 * when @bio is released. The caller must own @bio and is responsible for
1970 * synchronizing calls to this function.
1972 int bio_associate_blkcg(struct bio
*bio
, struct cgroup_subsys_state
*blkcg_css
)
1974 if (unlikely(bio
->bi_css
))
1977 bio
->bi_css
= blkcg_css
;
1982 * bio_associate_current - associate a bio with %current
1985 * Associate @bio with %current if it hasn't been associated yet. Block
1986 * layer will treat @bio as if it were issued by %current no matter which
1987 * task actually issues it.
1989 * This function takes an extra reference of @task's io_context and blkcg
1990 * which will be put when @bio is released. The caller must own @bio,
1991 * ensure %current->io_context exists, and is responsible for synchronizing
1992 * calls to this function.
1994 int bio_associate_current(struct bio
*bio
)
1996 struct io_context
*ioc
;
2001 ioc
= current
->io_context
;
2005 get_io_context_active(ioc
);
2007 bio
->bi_css
= task_get_css(current
, blkio_cgrp_id
);
2012 * bio_disassociate_task - undo bio_associate_current()
2015 void bio_disassociate_task(struct bio
*bio
)
2018 put_io_context(bio
->bi_ioc
);
2022 css_put(bio
->bi_css
);
2027 #endif /* CONFIG_BLK_CGROUP */
2029 static void __init
biovec_init_slabs(void)
2033 for (i
= 0; i
< BIOVEC_NR_POOLS
; i
++) {
2035 struct biovec_slab
*bvs
= bvec_slabs
+ i
;
2037 if (bvs
->nr_vecs
<= BIO_INLINE_VECS
) {
2042 size
= bvs
->nr_vecs
* sizeof(struct bio_vec
);
2043 bvs
->slab
= kmem_cache_create(bvs
->name
, size
, 0,
2044 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2048 static int __init
init_bio(void)
2052 bio_slabs
= kzalloc(bio_slab_max
* sizeof(struct bio_slab
), GFP_KERNEL
);
2054 panic("bio: can't allocate bios\n");
2056 bio_integrity_init();
2057 biovec_init_slabs();
2059 fs_bio_set
= bioset_create(BIO_POOL_SIZE
, 0);
2061 panic("bio: can't allocate bios\n");
2063 if (bioset_integrity_create(fs_bio_set
, BIO_POOL_SIZE
))
2064 panic("bio: can't create integrity pool\n");
2068 subsys_initcall(init_bio
);